Why Spotify AI more than music is secret to keeping listeners

Streaming music apps have been nudging users into the artificial intelligence era with a limited track record of success. But AI-based recommendation tools from Apple, Amazon and pure-play streaming company Spotify are moving ahead, with Spotify’s latest approach to the future of personal music discovery leaning into the AI prompt in multiple formats. Experts say these tech investments may be critical to Spotify’s ability to build a moat around its business as the core input, music, becomes commoditized across the streaming apps.

A new ChatGPT integration rolled out recently by Spotify allows users to connect their accounts directly to OpenAI’s generative AI chatbot. The launch is good for OpenAI in its broader effort to turn ChatGPT into a platform for third-party apps that function inside conversations. For Spotify, it’s a bet that personalized music and podcast recommendations will be improved through the now-familiar format of chatting with an AI and letting it know what you want.  

Spotify users can ask for songs, artists, albums, playlists or podcast episodes by mood, genre or topic. Results surface inside ChatGPT and open in the Spotify app for playback. Users can interact with the recommendation, and offer specificity beyond what is possible with a classic “like/dislike” feedback option. 

According to a Spotify spokesperson, the prompts are “an opportunity to uncover new tracks or revisit old favorites, or extend a ChatGPT conversation with a soundtrack that fits the moment.” 

Spotify said the integration is opt-in and that users can disconnect at any time. It also said it will not share music or podcast content with OpenAI for training purposes — addressing industry concerns around AI and copyrighted material.

Spotify also recently rolled out its Prompted Playlist feature inside the streaming music app, a “vibey” feature that allows users to tap into a feeling or memory in order to build a custom mix.

The rival streaming services connected to the big tech players are exploring similar AI features.

Apple has been gradually layering AI into Apple Music. The “Playlist Playground” beta feature is closest to what Spotify is doing, as it also focuses on chat-based AI interaction that allows users to tweak recommendations via chat. Apple recently introduced AutoMix, which uses AI to analyze songs and automatically blend tracks by matching tempo and beats, eliminating silence between songs, adding crossfades and so on. The company has also rolled out machine-learning tools such as lyric translation and pronunciation features. 

Amazon Music has offered a prompt-based playlist feature called Maestro since mid-2024, which allows listeners to generate playlists using text descriptions or even emojis. It remains in beta testing rather than full release.

Spotify executives have repeatedly described AI as central to the platform’s subscriber stickiness strategy. On a recent earnings call, leadership told investors that improvements in AI-driven discovery are central to keeping users engaged with the platform. “Our investments into personalization and AI are paying off,” said Alex Norström, co-chief executive officer. “It means people are spending more days in a month with us and across more moments,” he said.

Spotify’s interactive iDJ feature, introduced in 2023, has roughly 90 million subscribers using it as of the most recent earnings report, with users racking up over four billion hours of time spent on the app. Norström said Prompted Playlists as “instantly taken off with power users.”

“If iDJ is the chat interface to Spotify, where you can talk casually, Prompted Playlists is the Deep Research mode of Spotify,” he said. “It lets you describe and set rules for your own personalized playlists – literally writing your own algorithm. … There’s nothing else like it.”

Music catalogs commoditized, AI generates millions of songs

According to analysts who cover Spotify, the executive hype about AI may need to be the reality sooner rather than later for the company. Even though there are cases of musicians pulling music from specific apps from time to time as negative headline issues appear — it happened to Spotify most recently due to its founder and former CEO Daniel Ek’s defense tech investments — competitors including Apple Music, Amazon Music and YouTube Music offer largely overlapping catalogs and increasingly sophisticated recommendation engines.

“The catalogs at Amazon, Apple and YouTube are similar — nearly identical songs — to Spotify, just like Bing and Edge are nearly identical to Google,” said Michael Pachter, senior advisor, digital media, sports & entertainment at Wedbush Securities, who covered the streaming industry as research analyst for many years. (Wedbush has never had a rating on Spotify shares specifically.)

While Google’s search business faces its own AI threat, Pachter said it is also the best model for Spotify to look to in terms of how to maintain a user edge. “Google managed to widen its moat by offering a number of features that make the service stickier, including remembering my credit card and password info. I can’t even conceive of switching from Google Search, and I think that is what Spotify is trying to establish,” he said.

Switching costs may be subtle, but they can be significant. Users build libraries, curate playlists and train algorithms over years. Each additional integration, whether with a car dashboard, a voice assistant or now an AI chatbot — Spotify says it now connects to over 2,000 device types — can further entrench the ecosystem.

“I expect this ChatGPT integration will be widely used by Spotify users and wildly successful,” Pachter said. “Others may try to do the same thing, but the switching costs grow every time you make the effort to build your playlists on Spotify, and that’s what they are counting on,” he said.

Apple Music and other third-party apps do offer tools to export playlists when subscribers seek to change music services.

Others on Wall Street are less convinced than Pachter, but did come away from the most recent quarter more positive on the Spotify AI story and less worried about the risks it faces from AI music creation tools disrupting platforms like its own. Spotify’s stock price has slumped in the past year by close to 20%, though the stock has performed very strongly since its 2018 IPO.

“Spotify addressed this concern head‑on, arguing that AI supports rather than undermines its strategic position. By leaning into personalization, product innovation, and scale advantages, Spotify appears positioned to use AI to strengthen its platform, though the pace of adoption and industry alignment will remain key variables,” wrote Bank of America’s research team, which rates the shares a buy, in a February note after the most recent earnings.

Gustav Söderström, Spotify co-CEO, said during that earnings call that building a music app that users can talk to, and that fully understands each listener, will shift listening “from a passive experience to an interactive one.”

Mark Mulligan, managing director and analyst at MIDiA Research, a research firm that tracks the music market, says AI is going to be integral to streaming music behavior, but he is less convinced that the distinction between interactive and passive being made by Spotify is the likely result.

“Streaming music consumption has bifurcated between passive and active,” Mulligan said. “But this does not mean that the audience has split in two — everyone has. Even the most active music listeners spend more than half their time listening passively.”

In fact, he says the broader trend is toward more passive consumption through curated playlists, and also via features such as artist radio stations and AI DJs. “The direction of travel is towards more passive listening,” Mulligan said. Agentic features may represent a compromise, “a middle ground between passive and active listening,” he said. “It allows the user to expend a small amount of ‘lean forward’ effort in return for a large amount of ‘lean back’ listening.”

Typing a detailed prompt into ChatGPT may feel active, but Mulligan says “the more that the algorithm learns about the listener’s behavior and tastes, the better its recommendations get and therefore the less the user has to lean in, thus shifting the needle even further towards passive listening.”

In this AI interface-first model of streaming, the underlying content is important, but matters less to what ultimately makes the user feel rewarded. For example, an ability to explicitly exclude artists or narrow by subgenre could make AI-assisted discovery feel more tailored than traditional algorithmic playlists. If a listener enjoys some 1980s rock bands, such as Bon Jovi and Guns N’ Roses, but dislikes others from the same era, it’s easier to filter. Spotify can normally predict that if you like A and most people who like A also like B, then you’ll probably like B, but that doesn’t necessarily represent how user tastes are expressed. “With GPT, I could say ‘no Def Leppard’ and my lists would be scrubbed of them,” Pachter said.

Any predictions about the impact of AI on music, as with any industry, remain educated guesswork. But it is already very clear that AI is having an impact on the idea of a music catalog itself. According to a recent Rothschild & Co Redburn report, text-to-music platforms like Suno are reportedly generating around seven million songs per day, roughly equivalent to Spotify’s entire pre-AI catalogue every fortnight. “This is a deluge,” its analyst Ed Vyvyan stated.

Söderström hinted that it’s the dataset yet to come and not the deep tracks already in the stack that matter most to the future. “We are building a dataset that never existed,” he said on the recent earnings call. “We have had the song-to-song dataset, but no one had the language-to-song dataset. … You may think it is a canonical dataset, meaning there is a factual answer to, for example, what is workout music? There is no factual answer to what is workout music. … on average, for an American, it is usually hip hop. For a European, it is usually EDM. For many Scandinavians, it is something like heavy metal or even death metal. Then again, for a lot of Americans, millions at least, it is also death metal.”

“You cannot just have an LLM commoditize it as a fact, the way you can commoditize Wikipedia,” he said. “You actually need to have many, many hundreds of millions of listeners across the world’s markets constantly telling you what it means for that specific person.”